Fiscal adjustments by Steinar Holden Department of Economics University of Oslo Box 1095 Blindern 0317 Oslo Email: steinar.holden@econ.uio.no Homepage: http://folk.uio.no/~sholden/ First version: 27 January 2005 NB: Preliminary version Errors may occur Abstract Recent literature has argued that to reduce public debt permanently, a government has to cut public expenditure, specifically government employment and transfers. Tax increases supposedly give only a temporary reduction in public debt. This paper explores this conclusion using a new fiscal indicator, building on the indicator proposed by Braconier and Holden (1999). JEL Codes: E6, H6 Keywords: fiscal policy, fiscal indicators, budget indicators 1 Alesina and Perotti (1995) (AP) argue that • to reduce public debt permanently, a government has to cut public expenditure, specifically government employment and transfers, while tax increases don’t work “Permanent improvements are implemented mainly via cuts in two types of expenditure: transfer programmes and compensation of government employees. Temporary improvements are carried out almost exclusively via tax increases” This finding has received considerable attention in the literature, and also in textbooks, e.g. David Romer Advanced Macroeconomics, (2001), page 547. Ardagna, EER 2004, find smaller difference between tax increases and expenditures, but with a different method from AP, and without comparing explicitly. Is AP’s conclusion valid? 2 Alesina and Perottis’s analysis 19 OECD countries over the period 1960 - 92 Definition 1 Fiscal adjustment: Strong fiscal adjustment is a year with very tight fiscal stance, defined as a discretionary change in budget balance greater than 1.5 percent of GDP Discretionary change in budget balance is defined by use of Blanchard’s fiscal impulse; what the budget balance would have been if unemployment had remained the same as in the previous year. I.e. adjust the primary deficit (i.e. without net interest payments) for the effect of the change in unemployment. Similar results when using OECD structural budget balance instead of Blanchard’s fiscal impulse. 3 Definition 2 Successful adjustment: A successful adjustment in year t is defined as a very tight fiscal stance in year t such that the gross debt/GDP ratio in year t+3 is at least 5 percentage point of GDP lower than in year t. Focus on gross debt because • common in literature • Maastricht criteria • does not depend on valuation of government financial assets • but problem due to effect of sales of gov’t financial assets. 4 Table 11, AP Successful adjustment (14 obs) -2.74 (0.282) -2.19 (0.326) 0.44 (0.385) Unsuccessful adjustment (38 obs) -2.18 (0.101) -0.49 (0.188) 1.28 (0.181) Successful adjustment (14 obs) Expenditure -2.193 (0.326) Public investment -0.41 (0.089) Transfers -0.54 (0.183) Non-wage public -0.38 consumption (0.055) Government wages -0.58 (0.093) Subsidies -029 (0.211) Unsuccessful adjustment (38 obs) -0.49 (0.188) -0.26 (0.046) -0.02 (0.102) -0.09 (0.038) -0.07 (0.071) -0.08 (0.047) Fiscal impulse Expenditure Revenues Table 12, AP 5 This paper: Explore whether AP’s conclusion is valid, • using a novel fiscal indicator • a more sophisticated analysis (Frankly, another motivation for the study is to apply the novel indicator) The new fiscal indicator Decomposition of the change in the budget balance into • Discretionary changes are the effect of changes in the fiscal policy. • Induced changes arise as a consequence of changes in the economy; these are the changes that would take place even if fiscal policy were constant. In contrast, the OECD Structural budget balance distinguishes between • structural changes • cyclical changes 6 Revenues: unchanged fiscal policy implies that tax revenues are proportional to their respective tax bases. Distinguish between five different tax types. • Direct taxes on households (TD) are proportional to pre-tax household income (Hinc). • Direct taxes on the business sector (TB) are proportional to business income (Binc). • Social security contributions (TS) are proportional to the wage bill, excluding soc sec contributions (WL) • Indirect taxes (TI) are proportional to private consump. • Other revenues (TO) are proportional to GDP (Y). (1) ∆TIt ≡ TD,t-1 (Hinct/Hinct-1 –1)+ TB,t-1 (Binct/Binct-1 –1)+ T S,t-1(WLt/WLt-1 –1) + TI,t-1(Ct/Ct-1 –1) + TO,t-1(Yt/Yt-1-1) The discretionary change in revenues is defined as a residual (2) ∆TDt ≡ ∆Tt - ∆TIt, I.e. there is a discretionary increase in taxes if tax revenues increase above the growth in the tax bases. 7 For expenditures, excluding interest spending and capital transfers and other capital payments, unchanged policy is defined as • Public expenditures (excl. unemployment benefits) are proportional to trend GDP. ∆GIO,t ≡ GO,t-1 ( YTt /YTt-1 – 1) • Unemployment benefits, as a share of trend GDP, are proportional to the rate of unemployment. ∆GIU,t ≡ GU,t-1 [(Ut /Ut-1)( YTt /YTt-1) – 1] (trend real growth = average of real growth in past 10 years) Total induced change in government expenditure is (4) ∆GIt ≡ ∆GIO,t + ∆GIU,t Total discretionary change is a residual (5) ∆GDt ≡ ∆Gt – ∆GIt 8 The discretionary change in the budget balance (lower case denotes ratio to GDP) (6) ∆bDt ≡ ∆tDt - ∆gDt The overall budget balance is (7) Bt = Tt – Gt - iDt iDt is net interest payments The induced change in the budget balance, as a ratio to GDP, is defined as a residual (8) ∆bIt ≡ (B/Y)t - (B/Y)t-1 – ∆bDt. (implying that net interest payments are treated as induced) In contrast, the OECD structural budget balance adjusts the primary deficit for the effect of the output gap, i.e. the difference between actual and potential output. This implies that transfers and tax revenues are adjusted for changes in GDP. 9 Illustration of the motivation for the fiscal indicator: The crisis in Finland and indicators of fiscal stance Dramatic economic downturn in the early 1990s due to disruption of considerable exports to former Sovjet Union, as well as debt problems and banking crisis. Sharp reduction in GDP in 1991-92, while the fall in the main tax bases was lagging behind. The OECD indicator, the Structural Change Index (SCI) (which is the change in the Structural budget balance) “overadjusted” the budget balance for the reduction in GDP, and misleadingly indicated a neutral fiscal stance in 1992-94. In contrast, when the growth in the tax bases were lagging behind GDP-growth, 1993-95, the OECD indicator did not show a tight fiscal stance in 1995-97 10 GDP growth Tax base growth Unemployment FINLAND 0.2 0.15 0.1 0.05 0 -0.05 -0.1 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 FINLAND 0.04 0.03 0.02 0.01 0 dbd sci -0.01 -0.02 -0.03 -0.04 -0.05 2002 2001 2000 1999 1998 1997 1996 1995 1994 1993 1992 1991 1990 1989 1988 1987 1986 1985 1984 1983 1982 1981 -0.06 11 Returning to the comparison with AP: Evolution of debt as a ratio to GDP: Standard notation, D is debt, the total budget deficit is -B = ∆D = G – T + iD-1 (debt measured at the end of the year) ⎛D⎞ ⎛D⎞ ∆⎜ ⎟ = ⎜ ⎟ ⎝ Y ⎠t ⎝ Y ⎠t = D − Dt −1 Dt −1 Dt −1 ⎛D⎞ −⎜ ⎟ = t + − Y Y Y Yt −1 ⎝ ⎠t −1 t t ∆Dt Dt −1 Yt −1 Dt −1 ∆Dt ⎛ Yt −1 ⎞ ⎛ D ⎞ ∆Dt ∆Yt ⎛ D ⎞ + − = +⎜ − 1⎟ ⎜ ⎟ = − ⎜ ⎟ Yt Yt −1 Yt Yt −1 Yt Y Y Y Yt ⎝ Y ⎠t −1 t ⎝ t ⎠ ⎝ ⎠t −1 using ∆Dt/Yt = -Bt/Yt = ∆bIt + ∆bDt + (B/Y)t-1, (9) ∆(D/Y)t = ∆bIt + ∆bDt. - (B/Y)t-1 – (∆Y/Y)t (D/Y)t-1 , AP make a point that cyclical changes in the budget deficit are removed (corresponding to ∆bIt ), but they do not discuss the latter two components • lagged deficit (B/Y)t-1 and • growth-induced reduction in debt (∆Y/Y)t (D/Y)t-1 , 12 Our analysis: 18 OECD countries, 1970-2002, Gross debt Successful adjustment (20 obs) Fiscal impulse 0.025 (dbd) (.008) Expenditures (dgd) -0.009 (0.007) Revenues (dtd) 0.016 (0.010) -0.014 Lagged deficit (0.026) (B/Y)t-1 Growth-induc. red. 0.154 (0.063) (∆3Y/Y)t(D/Y)t-1 Lagged debt 0.640 (D/Y)t-1 (0.252) Unsuccessful adjustment (79 obs) 0.026 (0.015) -0.007 (0.010) 0.019 (0.012) -0.044 (0.049) 0.142 (0.098) 0.619 (0.308) We observe: • almost no difference in change in expenditures • almost no difference in change in revenues • large difference in lagged deficit (over a three year period, 3 percent higher deficit will lead to about 9 percent higher debt ratio) • no difference in initial debt nor in “growth-induced” reduction in debt Success of adjustment depends on the initial level of deficit Similar picture with net public debt; where lagged deficit differs by 5 percent (-0.001 and -0.051) 13 | country year fiscadj successg | |-----------------------------------------| 1. | Canada 1996 5 1| 2. | Canada 1997 5 1| 3. | Denmark 1986 5 1| 4. | Denmark 1999 5 1| 5. | Denmark 1985 5 1| |-----------------------------------------| 6. | Finland 1998 5 1| 7. | Finland 2000 5 1| 8. | Italy 1997 5 1| 9. | Netherlands 1972 5 1| 10. | Netherlands 1973 5 1| |-----------------------------------------| 11. | New Zealand 1993 5 1| 12. | New Zealand 2001 5 1| 13. | Norway 1996 5 1| 14. | Norway 1980 5 1| 15. | Norway 1995 5 1| |-----------------------------------------| 16. | Norway 1994 5 1| 17. | Sweden 1997 5 1| 18. | Sweden 1987 5 1| 19. | UK 1980 5 1| 20. | UK 1998 5 1| 14 Extended analysis, using all observations, not only the observations with very tight fiscal policy. First, considerable data problems in the sense that the budget identity does not hold. ∆(D/Y)t = - (B/Y)t - (∆Y/Y)t (D/Y)t-1 Regressing ∆(D/Y)t on (B/Y)t + (∆Y/Y)t (D/Y)t-1, we obtain (using gross debt) ∆(D/Y)t = 0.73 {(B/Y)t +(∆Y/Y)t (D/Y)t-1 }, R2 = 0.46 With net debt, we obtain ∆(D/Y)t = 0.95 {(B/Y)t +(∆Y/Y)t (D/Y)t-1 }, R2 = 0.51 Thus, the statistical discrepancy due to valuation-changes of financial assets, sales of financial assets, one-off events, etc, is considerable. 15 Panel data estimation: Dependent variable: The change in the gross debt ratio ∆3(D/Y)t+2 = (D/Y)t+2 - (D/Y)t-1 N = 443 Coef. Std. Err Coef. dgd 3.13 0.33 dgi 6.19 0.38 dtd -2.41 0.27 dti -3.46 0.43 (B/Y)t-1 -1.53 (∆3Y/Y)t(D/Y)t-1 Std. Err 2.01 0.45 -0.65 0.33 0.10 -1.08 0.12 -0.70 0.05 -0.21 0.05 _cons 0.01 0.01 0.03 0.01 _ R2 0.56 0.16 Some indication that discretionary change in expenditure has greater impact on the change in the debt ratio than the discretionary change in taxes has. 16 Dependent variable: The change in the net debt ratio ∆3(D/Y)t+2 = (D/Y)t+2 - (D/Y)t-1 N = 443 Coef. Std. Err Coef. dgd 3.71 0.35 dgi 6.05 0.39 dtd -3.39 0.28 dti -4.86 0.40 (B/Y)t-1 -2.08 (∆3Y/Y)t(D/Y)t-1 Std. Err 2.59 0.42 -1.42 0.32 0.11 -2.09 0.14 -0.56 0.04 -0.55 0.05 _cons 0.03 0.01 0.01 0.01 _ R2 0.59 0.35 What can explain the difference in effect of changes in expenditure and taxes? • Persistence in adjustment • Effect on GDP growth • Spurious correlation 17 Dependent variable: dgd Source | SS df MS Number of obs = 450 -------------+--------------------------F( 6, 443) = 11.53 Model | .00667 6 .00111 Prob > F = 0.0000 Residual | .04275 443 .000096 R-squared = 0.1351 -------------+------------------------------Adj R-squared = 0.1233 Total | .04942 449 .00011 Root MSE = .00982 -------------------------------------------------------------------------dgd | Coef. Std. Err. t –value ----------------------------------------------------dgd | L1 | .2481324 .0456562 5.43 L2 | -.0748581 .0452475 -1.65 dtd | L1 | -.0524612 .0333692 -1.57 (B/Y) | L1 | .0746334 .015686 4.76 (D/Y) | L1 | .0025336 .0018419 1.38 3 T (∆ Y /YT)| .0060944 .0029005 2.10 _cons | -.0001249 .0010112 -0.12 -------------------------------------------------------------------L1 denotes one lag Evidence that discretionary adjustment of expenditure is persistent 18 Dependent variable: dtd Source | SS df MS Number of obs = 465 -------------+------------------------------ F( 6, 458) = 8.50 Model | .0088 6 .001 Prob > F = 0.0000 Residual | .0797 458 .00017 R-squared = 0.1002 -------------+--------------------Adj R-squared = 0.0884 Total | .0886 464 .0001 Root MSE = .0132 -------------------------------------------------------------------------dtd | Coef. Std. Err. t-value ----------------------------------------------dgd | L1 | -.0168938 .0597263 -0.28 dtd | L1 | .1348477 .0447304 3.01 L2 | -.080253 .0438723 -1.83 (B/Y) | L1 | -.1033305 .0205852 -5.02 (D/Y) | L1 | -.0059312 .0023771 -2.50 3 T (∆ Y /YT) | .0083852 .0039142 2.14 _cons | -.0002403 .001329 -0.18 --------------------------------------------------------------------------L1 denotes one lag Evidence that discretionary adjustment of taxes is persistent, but less so than expenditure. 19 Dependent variable: dgdp (GDP – growth) Source | SS df MS Number of obs = 559 -------------+---------------------F( 5, 553) = 282.17 Model | .98141 5 .19 Prob > F = 0.0000 Residual | .3846 553 .000695 R-squared = 0.7184 -------------+------------------------Adj R-squared = 0.7159 Total | 1.366 558 .00244 Root MSE = .02637 --------------------------------------------------------------dgdp | Coef. Std. Err. t -values -------------+-------------------------------------------------------dtd | L1 | -.0822291 .0837724 -0.98 dgd | L1 | .3545363 .1124326 3.15 dgdp | L1 | .4112334 .0495375 8.30 L2 | -.2957682 .0491912 -6.01 3 T (∆ Y /YT) | .1883783 .0164863 11.43 _cons | .0172468 .0023748 7.26 -------------------------------------------------------------------------Does expenditure have a positive impact on GDP- growth, which would reduce the effect on debt ? 20 Dependent variable: dgdp (GDP – growth) Source | SS df MS Number of obs = 528 -------------+----------------------F( 9, 518) = 159.60 Model | .9514 9 .1057 Prob > F = 0.0000 Residual | .3430 518 .00066 R-squared = 0.7350 -------------+-----------------------------Adj R-squared = 0.7304 Total | 1.29452 527 .002 Root MSE = .02574 ----------------------------------------------------------------------dgdp | Coef. Std. Err. t-value -------------+-------------------------------------------------------dtd | -- | .2569406 .0850816 3.02 L1 | -.1365593 .0841232 -1.62 L2 | -.068974 .0844821 -0.82 dgd | -- | -.5445395 .1177279 -4.63 L1 | .5213867 .1214715 4.29 L2 | -.1277002 .1154573 -1.11 dgdp | L1 | .4613033 .0508272 9.08 L2 | -.2323622 .0521239 -4.46 3 T (∆ Y /YT) | .1623048 .0170007 9.55 _cons | .0150339 .0024467 6.14 Significant coefficients, but no clear effect of discretionary expenditure on GDP -growth 21 Dependent variable: (∆3Y/Y)t+2 (GDP –growth over 3 years) Source | SS df MS Number of obs = 547 -------------+---------------------F( 4, 542) = 248.01 Model | 9.6551 4 2.413 Prob > F = 0.0000 Residual | 5.275 542 .0097 R-squared = 0.6467 -------------+-----------------------Adj R-squared = 0.6441 Total | 14.9302 546 .0273 Root MSE = .09865 -------------------------------------------------------------------------F2.d3gdp | Coef. Std. Err. t - values -------------+-----------------------------------------------------------dgd | .1164163 .4213446 0.28 dtd | .3349165 .3141267 1.07 (B/Y) | L1 | .0739653 .111165 0.67 3 T (∆ Y /YT) | .7083434 .0234001 30.27 _cons | .0589786 .0085737 6.88 No evidence of impact on GPD-growth over the next three years. 22 Conclusions: • The success of fiscal adjustments depends on the level of deficit, not whether the adjustment is undertaken by increasing taxes or reducing expenditure. • Nevertheless some evidence that discretionary reduction in public expenditure has a greater impact on the evolution of public debt than discretionary tax increases have. • Some evidence that discretionary reduction of public expenditure is more persistent than discretionary increases in tax. • No evidence that discretionary changes in expenditure or taxes affect the average GDP-growth over the next three years. 23 . +-----------------------------------------+ | country year fiscadj successg | |-----------------------------------------| 1. | Canada 1996 5 1| 2. | Canada 1997 5 1| 3. | Denmark 1986 5 1| 4. | Denmark 1999 5 1| 5. | Denmark 1985 5 1| |-----------------------------------------| 6. | Finland 1998 5 1| 7. | Finland 2000 5 1| 8. | Italy 1997 5 1| 9. | Netherlands 1972 5 1| 10. | Netherlands 1973 5 1| |-----------------------------------------| 11. | New Zealand 1993 5 1| 12. | New Zealand 2001 5 1| 13. | Norway 1996 5 1| 14. | Norway 1980 5 1| 15. | Norway 1995 5 1| |-----------------------------------------| 16. | Norway 1994 5 1| 17. | Sweden 1997 5 1| 18. | Sweden 1987 5 1| 19. | UK 1980 5 1| 20. | UK 1998 5 1| |-----------------------------------------| 21. | Austria 1996 5 2| 22. | Austria 1997 5 2| 23. | Austria 1981 5 2| 24. | Austria 2001 5 2| 25. | Austria 1984 5 2| |-----------------------------------------| 26. | Belgium 1982 5 2| 27. | Belgium 1993 5 2| 28. | Belgium 1984 5 2| 29. | Belgium 1985 5 2| 30. | Belgium 1987 5 2| |-----------------------------------------| 31. | Belgium 1983 5 2| 32. | Belgium 1977 5 2| 33. | Canada 1986 5 2| 34. | Canada 1987 5 2| 35. | Canada 1995 5 2| |-----------------------------------------| 36. | Canada 1981 5 2| 37. | Canada 1994 5 2| 38. | Denmark 1981 . 2| 24 39. | Denmark 1984 5 2| 40. | Denmark 1983 5 2| |-----------------------------------------| 41. | Finland 1984 5 2| 42. | Finland 1993 5 2| 43. | Finland 1976 5 2| 44. | Finland 1994 5 2| 45. | Finland 1988 5 2| |-----------------------------------------| 46. | Finland 1981 5 2| 47. | France 1994 5 2| 48. | France 1987 5 2| 49. | France 1996 5 2| 50. | France 1979 5 2| |-----------------------------------------| 51. | France 1984 5 2| 52. | France 1983 5 2| 53. | Germany 1976 5 2| 54. | Germany 1994 5 2| 55. | Germany 1993 5 2| |-----------------------------------------| 56. | Germany 1996 5 2| 57. | Germany 1969 5 2| 58. | Germany 1982 5 2| 59. | Germany 1983 5 2| 60. | Germany 1966 5 2| |-----------------------------------------| 61. | Germany 1973 5 2| 62. | Germany 1997 5 2| 63. | Italy 1980 5 2| 64. | Italy 1982 5 2| 65. | Italy 1995 5 2| |-----------------------------------------| 66. | Italy 1967 5 2| 67. | Italy 1986 5 2| 68. | Italy 1976 5 2| 69. | Italy 1993 5 2| 70. | Italy 1990 5 2| |-----------------------------------------| 71. | Italy 1983 5 2| 72. | Japan 1990 5 2| 73. | Japan 1974 5 2| 74. | Japan 1984 5 2| 75. | Japan 1981 5 2| |-----------------------------------------| 76. | Netherlands 1982 5 2| 77. | Netherlands 1981 5 2| 78. | Netherlands 1983 5 2| 79. | Netherlands 1993 5 2| 80. | Netherlands 1991 5 2| 25 |-----------------------------------------| 81. | Norway 1999 5 2| 82. | Norway 1983 5 2| 83. | Norway 2000 5 2| 84. | Norway 1985 5 2| 85. | Spain 1996 5 2| |-----------------------------------------| 86. | Spain 1992 5 2| 87. | Spain 1994 5 2| 88. | Sweden 1995 5 2| 89. | Sweden 1971 5 2| 90. | Sweden 1994 5 2| |-----------------------------------------| 91. | Sweden 1996 5 2| 92. | Sweden 1976 5 2| 93. | Sweden 1984 5 2| 94. | Sweden 1975 5 2| 95. | Sweden 1989 5 2| |-----------------------------------------| 96. | Sweden 1983 5 2| 97. | UK 1997 5 2| 98. | UK 1981 5 2| 99. | UK 1977 5 2| 100. | US 1969 5 2| +-----------------------------------------+ 26 Blanchard’s fiscal impulse Estimate (TREND1 for 1960-75, TREND2 for 1976-92) (1) TRANSFt = α 0 + α1TREND1 + α 2TREND 2 + α 3U t + ε t Then calculate ) ) ) (2) TRANSFt (U t −1 ) = α 0 + αˆ1TREND1 + αˆ 2TREND 2 + α 3U t −1 + ε t Total primary expenditures, gt, equals tranfer expenditures TRANSF and non-transfer expenditures gNT,t; gt(Ut) = TRANSFt(Ut) + GNT,t. Consequently, non-transfer expenditures are assumed to be unaffected by the unemployment rate under a constant fiscal policy. By applying the same procedure to total revenues, tt, to obtain tt(Ut-1), we can construct the fiscal impulse as the difference between the unemployment-adjusted deficit and previous year’s deficit: (3) BFI t = ( g t (U t −1 ) − t t (U t −1 )) − ( g t −1 − t t −1 ) , 27